Abstract
The split triple-box girder, a newly adopted deck configuration for long-span rail–road dual-purpose bridges, is susceptible to vortex-induced vibration (VIV) due to its longitudinal slots, which significantly modify flow separation and vortex shedding. The presence of trains introduces additional aerodynamic interference, potentially altering the VIV behavior of the girder. Taking a long-span rail–road split triple-box girder as the prototype, this study combines sectional model wind tunnel tests with two-dimensional fluid–structure interaction simulations to investigate the influence of different train arrangements on vertical VIV responses and to elucidate the associated interference mechanisms. Two distinct vertical lock-in regions are identified. Train presence has a limited effect in the first lock-in region but markedly amplifies vibrations in the second. Among the investigated cases, a single windward train induces the largest vibration amplitudes, followed by a leeward train, while two parallel trains result in the smallest response. The aerodynamic interference is most pronounced at 0° and −3° angles of attack and weakens at +3°. Further analyses indicate that trains significantly modify surface pressure distributions by enhancing suction and pressure fluctuations over the middle and downstream boxes. Flow-field results show that different train arrangements alter vortex shedding patterns and wake development, thereby substantially affecting the vertical VIV characteristics of split triple-box girders.
Address
Han Yang: 1)Department of Bridge Engineering, Southwest Jiaotong University, Chengdu 610031, China
2)Key Laboratory for Wind Engineering of Sichuan Province, Chengdu 610031, China
Cunming Ma: 1)Department of Bridge Engineering, Southwest Jiaotong University, Chengdu 610031, China
2)Key Laboratory for Wind Engineering of Sichuan Province, Chengdu 610031, China
Shixiong Zheng: 1)Department of Bridge Engineering, Southwest Jiaotong University, Chengdu 610031, China
2)Key Laboratory for Wind Engineering of Sichuan Province, Chengdu 610031, China
Abstract
This study investigates the aerodynamic behavior of a train-bridge system on sea-crossing bridges under
cross-wind conditions based on 3D CFD. The effects of different wind speeds (Uw) and train speeds (Ut) on the
pressure distribution, vortex formation, and aerodynamic loads of the train-bridge system were analyzed. Results
show that cross-winds generate strong vortex structures near the head and tail of the train due to the shielding
interaction between the main girder and train. The aerodynamic loads of the train exhibit quasi-periodic variation
following a cosine law, with the head car showing drag reversal at high Uw. The bridge loads are primarily influenced
by Uw rather than Ut, with transient amplifications observed as the train passes. These research results offer crucial
theoretical support for the design and operation of the Train-Bridge System.
Key Words
aerodynamic characteristics; train-bridge system; cross-wind environments; train speed; flow
field
Address
Chen Guo: School of Civil Engineering and Architecture, Shaanxi University of Technology, Hanzhong, 723001, China
Youhao Liu: School of Civil Engineering and Architecture, Shaanxi University of Technology, Hanzhong, 723001, China
Guangling Guo: School of Civil Engineering and Architecture, Shaanxi University of Technology, Hanzhong, 723001, China
Abstract
Buffeting response prediction in long-span bridges has long resisted reliable solution, turbulent wind
loading, FEM uncertainties, and wind tunnel scaling limitations all compound one another, leaving a persistent gap
between model outputs and real-world behavior. The present work addresses this through a hybrid framework
combining similarity theory, FEM-based buffeting analysis, and a DeepRNN-LSTM framework, introduced to
correct discrepancies between scaled experimental results and full-scale bridge response.The LSTM operates as a
data-driven surrogate trained to minimize the residual between FEM predictions and measured responses, capturing
nonlinear temporal dependencies that conventional methods cannot adequately represent. Training and validation
drew on two distinct turbulent flow field conditions, pairing XNJD-3 wind tunnel measurements with FEM
simulations of the Balinghe Bridge 1:80 scale model, 1,088 m main span. After calibration, modal prediction errors
fell from over 8% to under 2%, with in-distribution buffeting errors of 0.24% MAPE and out-of-distribution RMS
error of 5.73%, consistent with independent physics-based benchmarks for the same bridge configuration, validated
on held-out test subsets via time-block partitioning across both flow fields. Uncertainty quantification confirmed
propagated prediction uncertainties remain below 4% across all response metrics.
Key Words
buffeting response; data-driven calibration; deep recurrent neural network; finite element model
updating; long short-term memory; long-span bridge; similarity theory; uncertainty quantification; wind
tunnel testing
Address
Mohamed Aljaboobi: School of Civil Engineering, Southwest Jiaotong University, Chengdu 610031, PR China
Shi-Xiong Zheng: School of Civil Engineering, Southwest Jiaotong University, Chengdu 610031, PR China
Muaadh Al-Qhfa: School of Civil Engineering, Southwest Jiaotong University, Chengdu 610031, PR China
Abstract
This study validates computational fluid dynamics (CFD) simulations of a generic train using OpenFOAM by comparing four Reynolds-averaged Navier-Stokes (RANS) turbulence models, which are Spalart-Allmaras, Shear Stress Transport (SST) k-ω, realizable k-ε and standard k-ε models. The objective is to assess their predictive capabilities in railway aerodynamics by comparing numerical results with previous experimental and numerical studies, focusing on drag coefficient, flow structures and pressure distribution. All models predict drag coefficient within 10% of reference values, except the standard k-ε, which substantially underpredicts the value. However, all models underestimate the drag coefficient relative to the experimental result. Analysis through streamline visualisation indicates that the Spalart-Allmaras model captures leading-edge vortex structures more consistently whereas SST k-ω and realizable k-ε produce larger vortices that remain attached further downstream. Conversely, standard k-ε incapable to capture the leading-edge vortex. The pressure coefficient (Cₚ) distribution along the midplane shows discrepancies at the front rooftop where all models predict a lower magnitude of negative pressure, suggesting a weaker suction effect. Despite minor discrepancies, the Spalart-Allmaras model shows a good agreement with the previous numerical study and proves to be the most computationally efficient, completing simulations faster than the other models.
Key Words
aerodynamic train; computational fluid dynamics; OpenFOAM; RANS; validation
Address
Nur Haziqah Shaharuddin: Department of Mechanical Engineering, Keio University, 3-14-1 Hiyoshi, Kohoku-Ku,
Yokohama 223-8522, Japan
Izuan Amin Ishak: Faculty of Engineering Technology, Universiti Tun Hussein Onn Malaysia,
Pagoh Higher Education Hub, Pagoh, 84600, Johor, Malaysia
Koji Fukagata:Department of Mechanical Engineering, Keio University, 3-14-1 Hiyoshi, Kohoku-Ku, Yokohama 223-8522, Japan
Abstract
This study focuses on identifying the most significant design parameters for a wind barrier that affect the aerodynamic performance of a high-speed train on a bridge subjected to crosswinds using statistical analysis. A fractional factorial design is applied to test four design factors, which are porosity, height, thickness, and hole diameter. The Reynolds-Averaged Navier-Stokes (RANS) with the Spalart-Allmaras turbulence model is utilised to perform simulations at a crosswind of 90° yaw angle to quantify the side force coefficient (Cₛ), which represents the dominant aerodynamic load acting laterally on the train. Design of experiment (DOE) and analysis of variance (ANOVA) are employed to determine the statistical significance of each factor. The simulation results show that porosity and height are the most significant parameters, by showing the p-value below the significance level of 0.05, with porosity having the highest F-value across the head, middle, and tail cars, indicating its dominant influence on crosswind performance. The reliability of the ANOVA results is confirmed by the main effects plots, which consistently identified porosity as the most influential factor. The findings suggest that, by optimising porosity and barrier height, the side loads on the train can be reduced effectively, thereby offering practical guidelines for the aerodynamic design of cost-efficient and high-performance wind barriers for high-speed train systems.
Key Words
ANOVA; crosswind; high-speed train; statistical analysis; wind barrier
Address
Nur Haziqah Shaharuddin: Department of Mechanical Engineering, Keio University, 3-14-1 Hiyoshi, Kohoku-Ku, Yokohama 223-8522, Japan
Koji Fukagata: Department of Mechanical Engineering, Keio University, 3-14-1 Hiyoshi, Kohoku-Ku, Yokohama 223-8522, Japan
Abstract
This paper analyses the wind-induced instabilities known as torsional galloping and divergence of single axis solar trackers using aeroelastic models. The onset of instability is influenced by the wind speed and direction, the tilt angle, and the structural properties of the tracker (i.e., torsional natural frequency, damping and stiffness). Trackers at tilt angles α=15° to 25° are most susceptible to torsional galloping. For tilt angles, α<5°, these trackers become unstable at a higher reduced wind speed, Ured, due to chained instability where torsional divergence transitions to torsional galloping. Trackers at tilt angle α>40° are less likely to experience instability. Torsional galloping is initiated when the aerodynamic damping counteracts the structural damping, and the tracker loses the ability to resist the fluctuating moments acting on the tube. The critical wind speed at which instability is initiated can be estimated using the aerodynamic derivative, A₂*, derived more accurately from the autocorrelation of the moment, which accounts for the 'fluid memory' influencing the moments on the torque tube. Employing a target maximum Ured can ensure stable operation by keeping the tracker within a range where aerodynamic damping contributes positively to the effective damping.
Key Words
aeroelastic instability; divergence; galloping; solar tracker; turbulence
Address
Juan I. Zaracho:Cyclone Testing Station, James Cook University, Townsville, QLD, Australia
John D. Ginger:Cyclone Testing Station, James Cook University, Townsville, QLD, Australia
John D. Holmes:JDH Consulting, Mentone, VIC, Australia
Korah Parackal:Insurance Institute for Business & Home Safety, Richburg, SC, USA
open access